Species level quantitative metagenomics Preprocessing and ... · 36626 - Next Generation Sequencing Analysis Preprocessing and SNP calling Natasja S. Ehlers, PhD student Center for

36626 - Next Generation Sequencing Analysis

Preprocessing and SNP calling Natasja S. Ehlers, PhD student Center for Biological Sequence Analysis Functional Human Variation Group

Species level quantitative metagenomics

Simon Rasmussen36626: Next Generation Sequencing analysis

DTU Bioinformatics


Nothing really new hereNothing new- but the technology


Classical measures

• Abundance

• Richness

• Rarefaction

• Diversity

Global terrestrial diversity

Abundance (counts)


Species richness

• The number of different species in a system

13/06/2016 Quantitative Metagenomics 10 DTU Sytems Biology, Technical University of Denmark

Abundance (Count)

Lion 64 Zebra 128 Giraffe 64 leopard 64 rhinoceros 64 hippopotamus 128 gazelle 128 elephant 64 monkey 9

9 observed species


Rarefaction

• Species richness is a function of our no. observations

• When have we sampled enough?

Rarefaction curve


Species diversity• “Effective number of species” represented in a

dataset

• Richness & Evenness

• Alpha diversity (within sample)

• Beta diversity (between samples)


Alpha diversity• Shannon index

• Quantify the entropy (information content)

• Quantifies the uncertainty (degree of surprise) associated with a prediction

Pi = species proportionR = observed species


Alpha diversity


Richness

Lion 1 Zebra 2 Giraffe 1 Leopard 1 Rhinoceros 1 Hippopotamus 2 Gazelle 2 Elephant 1 Monkey 0

Species richness estimators: Chao1 index = Sobs + f12/(2f2) Sobs = observed species f1 = species observed once f2 = species observed twice

8 observed species

Chao1 index = 8 + 52/(2*3) = 12.17


Richness

Lion 1 Zebra 2 Giraffe 1 Leopard 1 Rhinoceros 1 Hippopotamus 2 Gazelle 2 Elephant 1 Monkey 0

Species richness estimators: Chao1 index = Sobs + f12/(2f2) Sobs = observed species f1 = species observed once f2 = species observed twice

8 observed species

Chao1 index = 8 + 52/(2*3) = 12.17

H’ = -(ln(0.090.09) + ln(0.180.18) + … = 2.0


Beta diversity• Between samples (communities)


Beta-Diversity

Lion 0 2 Zebra 3 2 Giraffe 0 4 Leopard 0 2 Rhinoceros 1 2 Hippodrome 4 0 Gazelle 0 1 Elephant 1 0 Total 9 13


Beta-Diversity

Lion 0 2 Zebra 3 2 Giraffe 0 4 Leopard 0 2 Rhinoceros 1 2 Hippodrome 4 0 Gazelle 0 1 Elephant 1 0 Total 9 13


Bray-curtis dissimilarity


Beta-Diversity Lion 0 2 Zebra 3 2 Giraffe 0 4 Leopard 0 2 Rhinoceros 1 2 Hippodrome 4 0 Gazelle 0 1 Elephant 1 0 Total 9 13

Bray-Curtis dissimilarity metric

Bij = 1 - 2Cij / (Si + Sj) C = sum of the lowest count of common species S = total count of the sample Bs1s2 = 1 – 2*3 / 22 = 0.73 - Dissimilar C = 3 Ss1 + Ss2 = 22

0 ≤ B ≤ 1 Bij = 1 - 2Cij / (Si + Sj)

C = sum of the lowest count of common species

S = total count of the sample

Bs1s2 = 1 - 2*(2+1) / (9 + 13) = 0.73

0 ≤ B ≤ 1







0 ≤ B ≤ 1 Bij = 1 - 2Cij / (Si + Sj)



Bs1s2 = 1 - 2*(2+1) / (9 + 13) = 0.73

0 ≤ B ≤ 1







0 ≤ B ≤ 1 Bij = 1 - 2Cij / (Si + Sj)



Bs1s2 = 1 - 2*(2+1) / (9 + 13) = 0.73

0 ≤ B ≤ 1


Sampling effect• To be fair we should sample equally in the systems

we investigate

downsizingTo be fair we should sample equally carefully in the systems we investigate

Did you find any bacteria?

No - not here

Sample sizes


Sample sizes

• Accounting for different sample sizes:

• Normalise to sample size

• Rarefy (downsize) samples

• Statistically model the variance


Normalising


Sample Sizes

Lion 64 1 Zebra 128 2 Giraffe 64 1 Leopard 64 1 Rhinoceros 64 1 Hippopotamus 128 2 Gazelle 128 2 Elephant 64 1 Monkey 9 0 Total 713 11

Normalize to library size: Norm = ni/ntot

Lion 8.98 9.09 Zebra 17.95 18.18 Giraffe 8.98 9.09 Leopard 8.98 9.09 Rhinoceros 8.98 9.09 Hippopotamus 17.95 18.18 Gazelle 17.95 18.18 Elephant 8.98 9.09 Monkey 1.26 0 Total 100 100

N = ni/ntot

Issue with different sampling power (higher chance of observing rare species)


Downsize / rarefy

• Resample an equal number of observations (reads) from each sample

• Select the target depth carefully

• The more reads we keep the more sensitive

• We may have to remove samples with few counts


Downsize / rarefy


Sample Sizes

Rarefying to smaller library size:



Resample x amount of observations


Sample sizes• Statistically model the variance &

heteroscedasticity

• Use packages developed for RNA-seq such as DESeq2 and edgeR (negative binomial)

• Wilcoxon rank-sum statistics (ok for downsized data)

Lets try itTake a sample and count the animals!

Determine Richness, Species abundance and calculate Shannon diversity index

Calculate Bray-Curtis dissimilarity with your neighbour


Computational exercise• Continuation from yesterday

• 124 gut microbiome samples (MetaHIT)

• Sequenced using Illumina (~5Gbp)

• De novo assembly, gene prediction, gene count matrix

Species level quantitative metagenomics Preprocessing and ... · 36626 - Next Generation Sequencing Analysis Preprocessing and SNP calling Natasja S. Ehlers, PhD student Center for

Documents